by Paul Ellerman, Director of Reliability pellerman@microsemi.com Calculating Reliability using FIT & MTTF: Arrhenius HTOL Model Scope Establish a method for calculating the standard reliability values Failure Rate (), Failures in Time (FIT) and Mean Time to Failure (MTTF) using the Arrhenius High Temperature Operating Life (HTOL) model. The focus of this paper is to present the applicable equations, terms and definitions along with an example of an Excel driven reliability calculator used to perform these calculations. Terms & Definitions Reliability is defined as the probability that a component or system will continue to perform its intended function under stated operating conditions over a specified period of time. The reliability level is derived by monitoring the functional stability of a number of representative subjects operating under elevated stress conditions resulting in a statistical prediction of reliability. Two approaches to establishing a reliability level is to evaluate either the probability of survival or the probability of failure. Either method is equally effective, but the most common method is to calculate the probability of failure or Rate of Failure (). The values most commonly used when calculating the level of reliability are FIT (Failures in Time) and MTTF (Mean Time to Failure) or MTBF (Mean Time between Failures) depending on type of component or system being evaluated. Equations & Calculations Failure Rate () in this model is calculated by dividing the total number of failures or rejects by the cumulative time of operation. In the HTOL model, the cumulative time of operation is referred to as Equivalent Device Hours (EDH): EDH D H A f D Number of Devices Tested H Test Hours per Device A f Acceleration Factor derived from the Arrhenius equation 1 of 6
The Failure Rate () per hour is shown by: r D H A hour f r EDH r number of failures or rejects. Acceleration Factor (A f ) is the test time multiplier derived from the Arrhenius equation. This equation calculates the time acceleration value that results from operating a device at an elevated temperature. The test type used to achieve this is generally referred to as High Temperature Operating Life (HTOL) or Burnin. More specific terms for these tests depend on the type of technology under test. Tests such as High Temperature Reverse Bias (HTRB), DC Burn-in and Alternating Current Operating Life (ACOL) are technology specific tests within the HTOL category. The Acceleration Factor (A f ) is shown by: E 1 k T 1 T a use test A f e E a Activation Energy (ev) of the failure mode k (Boltzmann's Constant) 8.617 x 10-5 ev/ K T use Use Temperature (standardized at 55 C or 38 K) T test Test Temperature (HTOL temperature in K) K (degrees Kelvin) 73 + C ev electron volts Activation energy (E a ) is an empirical value that is the minimum energy required to initiate a specific type of failure mode that can occur within a technology type. Oxide defects, bulk silicon defects, mask defects, electro-migration and contamination are some examples of such failure modes, each with an unique associated E a. In lieu of using empirical data for each individual failure mode, it is generally accepted that a standard single value of E a 0.7 ev provides a reasonably accurate average E a value for diode type semiconductors. To derive a more statistically accurate calculation for Failure Rate (), the number of rejects (r) is replaced with the probability function using Chi-squared (X ). r ~ Χ ( α, ν ) X / (Chi-squared/) is the probability estimation for the number of failures or rejects. α (alpha), confidence level (CL) or probability, is the applicable percent area under the X probability distribution curve; reliability of 6
calculations use α 0.6 (or 60%). Note: The total area under the X curve is always 1, so α 1 (or 100%). ν (nu), degrees of freedom (DF), determines the shape of the X curve; reliability calculations use ν r + where r number of failures or rejects. The equation for Failure Rate () per hour then becomes: hour ( α, ν ) Χ ( α, ν ) Χ D H A f EDH FIT (Failures in Time) is a standard industry value defined as the Failure Rate () per billion hours. FIT FIT hours 9 10 MTTF (Mean Time to Failure or θ) is another standard industry value which provides the average time to failure of Non-repairable Items such as light bulbs and diodes or unserviceable systems such as satellites or other unmanned space craft. For items with long life expectancies, it is often a more useful to report MTTF in years rather than hours. 1 MTTF hours or MTTF years ( ) hours hours 4 365 1 MTBF (Mean Time between Failures) is used to describe Repairable Items such as compressors and aircraft. MTBF uses MTTF as one factor and Mean Time to Repair (MTTR) as the other to capture the complete break-down and repair cycle. The primary purpose of MTBF is to identify appropriate preventive maintenance schedules to avoid, perhaps indefinitely, catastrophic failures due to predictable piece part wear-out. As a rule of thumb, component reliability centers around MTTF since most components cannot be repaired. MTBF is shown by: MTBF MTTF + MTTR MTTR is the Mean Time to Repair. Reliability Calculation Tools Over the years, there have been several methods used to derive the value for X (Chisquared); everything from the use of probability tables to the application of 3 of 6
approximation equations. More detail on this subject is discussed in Calculating Chisquared (X ) for Reliability Equations by this author. The following is an excerpt. In MS Excel 003 and later versions, the CHIINV function calculates X from the input data probability and deg_freedom rendering accurate values throughout the entire use range up to 4-decimal places. This method is currently used as a standard in the statistics community for calculating values for the X distribution. In Excel, the formula is expressed as: CHIINV(probability,deg_freedom) In determining ν (nu), degrees of freedom or deg_freedom, we use the equation described previously where r number of failures or rejects: ( r + 1) ν r + Since reliability calculations require the left portion of area under the X distribution curve and CHIINV happens to calculate the right side, we must use the equation below to determine the correct value for probability: CHIINV ( probability) 1 α when using decimal values for α, or CHIINV ( probability) 1 α 100 when using percentage values for α. Example: for α.60 or 60%, CHIINV(probability).40 or 40% respectively. Figure 1 shows an example of an Excel driven reliability calculator which greatly simplifies the calculation activity. 4 of 6
5 of 6
Reference Material 1. Patrick D. T. O Connor, Practical Reliability Engineering, 4 th ed., John Wiley & Sons, UK, (010). Charles E. Ebeling, An Introduction to Reliability and Maintainability Engineering, nd ed., Waveland Press, USA (010) 3. Wayne B. Nelson, Accelerated Testing-Statistical Models, Test Plans & Data Analysis, John Wiley & Sons, USA (004) 4. Dimitri Kececioglu & Feng-Bin Sun, Burn-In Testing-Its Quantification & Optimization, DEStech Publications, USA (003) 5. Samuel M. Selby, Ph.D. Sc.D., Standard Mathematical Tables, 1th ed., The Chemical Rubber Co. (CRC), Cleveland, Ohio, USA (197) 6. Mary Gibbons Natrella, Experimental Statistics, National Bureau of Standards Handbook 91, United States Department of Commerce, Washington D.C., USA (1963) 7. Thomas Pyzdek, The Six Sigma Handbook, McGraw-Hill, New York, USA (003) 8. Ray DiBugnara, Model for Calculating Part Reliability from Life Test Data, Microsemi Corporation (009) 9. Paul Ellerman, Calculating Chi-squared (X ) for Reliability Equations, MicroNote 1003, Microsemi Corporation (01) 6 of 6